U.S. patent application number 14/824070 was filed with the patent office on 2016-02-11 for multi-view fingerprint matching.
The applicant listed for this patent is Synaptics Incorporated. Invention is credited to Rohini Krishnapura, Anthony P. Russo.
Application Number | 20160042247 14/824070 |
Document ID | / |
Family ID | 55267644 |
Filed Date | 2016-02-11 |
United States Patent
Application |
20160042247 |
Kind Code |
A1 |
Russo; Anthony P. ; et
al. |
February 11, 2016 |
MULTI-VIEW FINGERPRINT MATCHING
Abstract
A method and a device are provided for performing a recognition
process. The recognition process compares an individual fingerprint
view to a fingerprint enrollment template in order to determine
whether a match has been found. The determination of a match is
based on individual match statistics collected between the
individual fingerprint view and each view of the fingerprint
enrollment template. Additionally, inter-view match statistics
between each view of the fingerprint enrollment template may also
be determined. The inter-view match statistics can be analyzed
along with the individual match statistics to further inform the
determination of a match between the individual fingerprint view
and the fingerprint enrollment template.
Inventors: |
Russo; Anthony P.; (New
York, NY) ; Krishnapura; Rohini; (San Jose,
CA) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
Synaptics Incorporated |
San Jose |
CA |
US |
|
|
Family ID: |
55267644 |
Appl. No.: |
14/824070 |
Filed: |
August 11, 2015 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
62036037 |
Aug 11, 2014 |
|
|
|
Current U.S.
Class: |
382/125 |
Current CPC
Class: |
G06K 9/6202 20130101;
G06K 9/00093 20130101; G06K 9/00087 20130101; G06K 9/00026
20130101; G06K 9/52 20130101; G06K 9/66 20130101 |
International
Class: |
G06K 9/62 20060101
G06K009/62; G06K 9/66 20060101 G06K009/66; G06K 9/00 20060101
G06K009/00 |
Claims
1. A method of biometric matching to an enrollment template, the
method comprising: acquiring a verification template, the
verification template comprising a verification view of a biometric
sample captured by an input device; comparing the verification view
to a plurality of individual enrollment views of the enrollment
template to determine individual match statistics between the
verification view and the individual enrollment views; calculating
a composite match score between the verification template and the
enrollment template as a function of the individual match
statistics and of inter-view match statistics between at least one
pairing of the individual enrollment views within the enrollment
template; comparing the composite match score to a threshold; and
indicating a biometric match between the verification template and
the enrollment template if the composite match score satisfies the
threshold.
2. The method of claim 1, wherein calculating the composite match
score includes comparing the individual match statistics to the
inter-view match statistics.
3. The method of claim 2, wherein calculating the composite match
score includes an adjustment based on correspondence between the
individual match statistics and the inter-view match
statistics.
4. The method of claim 1, wherein calculating the composite match
score includes inputting the individual match statistics and the
inter-view match statistics into a machine learning classifier.
5. The method of claim 4, wherein the individual match statistics
include individual geometric transformations and individual match
scores between the verification view and each of the individual
enrollment views, wherein the inter-view match statistics include
at least one inter-view geometric transformation for the at least
one pairing of enrollment views, and wherein the individual
geometric transformations, the individual match scores, and the
inter-view geometric transformation are input into the machine
learning classifier to calculate the composite match score.
6. The method of claim 1, wherein the individual match statistics
include individual match scores between the verification view and
each of the individual enrollment views, wherein calculating the
composite match score includes combining the individual match
scores with an adjustment based on the inter-view match
statistics.
7. The method of claim 1, wherein the individual match statistics
include individual geometric transformations determined from
comparing the verification view to each of the individual
enrollment views, wherein the inter-view match statistics include
at least one inter-view geometric transformation for the at least
one pairing of enrollment views, and wherein calculating the
composite match score includes comparing the individual geometric
transformation with at least one derived transformation for the
verification view, wherein the at least one derived transformation
is derived from the at least one inter-view geometric
transformation.
8. The method of claim 1, wherein the individual match statistics
and the inter-view match statistics each include a number of
matched minutia.
9. A device for biometric matching to an enrollment template, the
device comprising: a processing system configured to: acquire a
verification template, the verification template comprising a
verification view of a biometric sample captured by an input
device; compare the verification view to a plurality of individual
enrollment views of the enrollment template to determine individual
match statistics between the verification view and the individual
enrollment views; calculate a composite match score between the
verification template and the enrollment template as a function of
the individual match statistics and of inter-view match statistics
between at least one pairing of the individual enrollment views
within the enrollment template; compare the composite match score
to a threshold; and indicate a biometric match between the
verification template and the enrollment template if the composite
match score satisfies the threshold.
10. The device of claim 9, wherein calculating the composite match
score includes comparing the individual match statistics to the
inter-view match statistics.
11. The device of claim 10, wherein calculating the composite match
score includes an adjustment based on correspondence between the
individual match statistics and the inter-view match
statistics.
12. The device of claim 9, wherein calculating the composite match
score includes inputting the individual match statistics and the
inter-view match statistics into a machine learning classifier.
13. The device of claim 12, wherein the individual match statistics
include individual geometric transformations and individual match
scores between the verification view and each of the individual
enrollment views, wherein the inter-view match statistics include
at least one inter-view geometric transformation for the at least
one pairing of enrollment views, and wherein the individual
geometric transformations, the individual match scores, and the
inter-view geometric transformation are input into the machine
learning classifier to calculate the composite match score.
14. The device of claim 9, wherein the individual match statistics
include individual match scores between the verification view and
each of the individual enrollment views, wherein calculating the
composite match score includes combining the individual match
scores with an adjustment based on the inter-view match
statistics.
15. The device of claim 9, wherein the individual match statistics
include individual geometric transformations determined from
comparing the verification view to each of the individual
enrollment views, wherein the inter-view match statistics include
at least one inter-view geometric transformation for the at least
one pairing of enrollment views, and wherein calculating the
composite match score includes comparing the individual geometric
transformation with at least one derived transformation for the
verification view, wherein the at least one derived transformation
is derived from the at least one inter-view geometric
transformation.
16. A device for fingerprint matching to an enrollment template,
the device comprising: a fingerprint sensor; and a processing
system configured to: acquire a verification template, the
verification template comprising a verification view of a
fingerprint sample captured by the fingerprint sensor; compare the
verification view to a plurality of individual enrollment views of
the enrollment template to determine individual match statistics
between the verification view and the individual enrollment views;
calculate a composite match score between the verification template
and the enrollment template as a function of the individual match
statistics and of inter-view match statistics between at least one
pairing of the individual enrollment views within the enrollment
template; compare the composite match score to a threshold; and
indicate a fingerprint match between the verification template and
the enrollment template if the composite match score satisfies the
threshold.
17. The device of claim 16, wherein calculating the composite match
score includes comparing the individual match statistics to the
inter-view match statistics.
18. The device of claim 16, wherein calculating the composite match
score includes inputting the individual match statistics and the
inter-view match statistics into a machine learning classifier.
19. The device of claim 18, wherein the individual match statistics
include individual geometric transformations and individual match
scores between the verification view and each of the individual
enrollment views, wherein the inter-view match statistics include
at least one inter-view geometric transformation for the at least
one pairing of enrollment views, and wherein the individual
geometric transformations, the individual match scores, and the
inter-view geometric transformation are input into the machine
learning classifier to calculate the composite match score.
20. The device of claim 16, wherein the individual match statistics
include individual match scores between the verification view and
each of the individual enrollment views, wherein calculating the
composite match score includes combining the individual match
scores with an adjustment based on the inter-view match statistics.
Description
CROSS REFERENCE TO RELATED APPLICATIONS
[0001] The application claims the benefit of U.S. Provisional
Patent Application Ser. No. 62/036,037, filed on Aug. 11, 2014.
FIELD OF THE INVENTION
[0002] This disclosure generally relates to electronic devices, and
more particularly to electronic devices configured to perform a
biometric recognition process.
BACKGROUND OF THE INVENTION
[0003] Biometric recognition systems are used for authenticating
and/or verifying users of devices incorporating the recognition
systems. Biometric sensing technology provides a reliable,
non-intrusive way to verify individual identity for recognition
purposes.
[0004] Fingerprints, like various other biometric characteristics,
are based on unalterable personal characteristics and thus are a
reliable mechanism to identify an individual. There are many
potential applications for utilization of fingerprint sensors. For
example, fingerprint sensors may be used to provide access control
in stationary applications, such as security checkpoints.
Electronic fingerprint sensors may also be used to provide access
control in portable applications, such as portable computers,
personal data assistants (PDAs), cell phones, gaming devices,
navigation devices, information appliances, data storage devices,
and the like. Accordingly, some applications, in particular
applications related to portable devices, may require recognition
systems that are both small in size and highly reliable.
[0005] Sometimes, the sensor may only be large enough to capture a
partial view of the biometric sample being sensed. For example, a
partial fingerprint sensor will only be large enough to capture a
partial image of a user's fingerprint. This can present several
challenges when attempting to reliably recognize the pattern
against a stored enrollment template using only the partial view.
For one, this partial view provides less discriminative information
for the matching system to utilize when attempting to reliably
recognize the fingerprint against a stored enrollment template.
Additionally, system may need to account for the user presenting
different portions of the same fingerprint in different match
attempts.
[0006] An enrollment template derived from multiple views of the
enrolled fingerprint provides a possible solution, but attempting
to perform matching between a candidate verification view and a
larger enrollment template that is derived from multiple enrollment
views is a challenging task.
[0007] One possible solution is to stitch together the multiple
enrollment views into a single larger view. However, achieving a
perfect alignment between the views is often not feasible, and
inaccuracies in these alignments can produce distortions at
boundaries and overlapping regions between the views that can
detrimentally impact match performance.
[0008] Another possible solution is to store the enrollment views
separately in the enrolled template, and compute individual match
scores with each of the views separately. The overall score between
the templates can then be based on a summation or other simple
combination of the scores to the individual views. Geometric
relationships between the enrollment views can be used to constrain
the alignments from the verify view to each enrollment view in
order to simplify the computation of each individual match score.
However, this can cause false results because the overall score is
a function of match scores to only the individual views. For
example, the combination of individual scores can cause a false
match when an imposter matches strongly with only a single view or
small subset of views, and conversely, the combination of
individual scores can cause a false non-match when a true user
matches only weakly with several of the views.
[0009] In view of the above, there is a need for a recognition
system that can provide a highly reliable recognition process based
on a partial view or views of a user's fingerprint. Embodiments of
the disclosure provide such a highly reliable recognition system
for performing a recognition process based on the partial view or
views of the user's fingerprint. These and other advantages of the
disclosure, as well as additional inventive features, will be
apparent from the description of the disclosure provided
herein.
BRIEF SUMMARY OF THE INVENTION
[0010] One embodiment provides a method of biometric matching to an
enrollment template. The method includes acquiring a verification
template, the verification template including a verification view
of a biometric sample captured by an input device. The method
further includes comparing the verification view to a plurality of
individual enrollment views of the enrollment template to determine
individual match statistics between the verification view and the
individual enrollment views. A composite match score between the
verification template and the enrollment template is calculated as
a function of the individual match statistics and of inter-view
match statistics between at least one pairing of the individual
enrollment views within the enrollment template. The composite
match score is compared to a threshold, and a biometric match
between the verification template and the enrollment template is
indicated if the composite match score satisfies the threshold.
[0011] Another embodiment includes a device for biometric matching
to an enrollment template. The device includes a processing system
configured to acquire a verification template, the verification
template including a verification view of a biometric sample
captured by an input device. The processing system is further
configured to compare the verification view to a plurality of
individual enrollment views of the enrollment template to determine
individual match statistics between the verification view and the
individual enrollment views. The processing system is configured to
calculate a composite match score between the verification template
and the enrollment template as a function of the individual match
statistics and of inter-view match statistics between at least one
pairing of the individual enrollment views within the enrollment
template. The processing system is configured to compare the
composite match score to a threshold, and a biometric match between
the verification template and the enrollment template is indicated
if the composite match score satisfies the threshold.
[0012] Another embodiment includes a device for fingerprint
matching to an enrollment template. The device includes a
fingerprint sensor. and a processing system. The processing system
is configured to acquire a verification template, the verification
template including a verification view of a fingerprint sample
captured by the fingerprint sensor. The processing system is
further configured to compare the verification view to a plurality
of individual enrollment views of the enrollment template to
determine individual match statistics between the verification view
and the individual enrollment views. The processing system is
configured to calculate a composite match score between the
verification template and the enrollment template as a function of
the individual match statistics and of inter-view match statistics
between at least one pairing of the individual enrollment views
within the enrollment template. The processing system is configured
to compare the composite match score to a threshold, and a
fingerprint match between the verification template and the
enrollment template is indicated if the composite match score
satisfies the threshold.
BRIEF DESCRIPTION OF THE DRAWING(S)
[0013] The accompanying drawings incorporated in and forming a part
of the specification illustrate several aspects of the present
invention and, together with the description, serve to explain the
principles of the invention. In the drawings:
[0014] FIG. 1 is a block diagram of an exemplary device that
includes an input device and a processing system, in accordance
with an embodiment of the invention;
[0015] FIG. 2a is an image of a fingerprint;
[0016] FIG. 2b is an enhanced image of the fingerprint of FIG.
2a;
[0017] FIG. 3 is an illustration of various types of minutiae
points of a fingerprint;
[0018] FIG. 4 is a block diagram of a matcher from the device of
FIG. 1, in accordance with an embodiment of the invention;
[0019] FIG. 5 is a schematic diagram of possible fingerprint views
for use by the matcher of FIG. 4, in accordance with a particular
embodiment of the invention;
[0020] FIG. 6 is graphical representation of match statistics
between a verification view and a plurality of enrollment views
from an enrollment template, in accordance with an embodiment of
the invention;
[0021] FIGS. 7(a)-7(c) are schematic diagrams of a recognition
process, in accordance with an embodiment of the invention;
[0022] FIG. 8 is a schematic diagram of fingerprint views and
associated geometric transformations, in accordance with an
embodiment of the invention; and
[0023] FIG. 9 is a flow chart for a recognition process performed
by the matcher of FIG. 4, in accordance with an embodiment of the
invention.
[0024] While the disclosure will be described in connection with
certain preferred embodiments, there is no intent to limit it to
those embodiments. On the contrary, the intent is to cover all
alternatives, modifications and equivalents as included within the
spirit and scope of the disclosure as defined by the appended
claims.
DETAILED DESCRIPTION OF THE INVENTION
[0025] The following detailed description is merely exemplary in
nature and is not intended to limit the invention or the
application and uses of the invention. Furthermore, there is no
intention to be bound by any expressed or implied theory presented
in the preceding technical field, background, brief summary or the
following detailed description.
[0026] Various embodiments of the present invention provide input
devices and methods that facilitate improved usability.
[0027] Turning now to the figures, FIG. 1 is a block diagram of an
electronic system or device 100 that includes an input device such
as sensor 102 and processing system 104, in accordance with an
embodiment of the invention. As used in this document, the term
"electronic system" (or "electronic device") broadly refers to any
system capable of electronically processing information. Some
non-limiting examples of electronic systems include personal
computers of all sizes and shapes, such as desktop computers,
laptop computers, netbook computers, tablets, web browsers, e-book
readers, and personal digital assistants (PDAs). Additional example
electronic devices include composite input devices, such as
physical keyboards and separate joysticks or key switches. Further
example electronic systems include peripherals such as data input
devices (including remote controls and mice), and data output
devices (including display screens and printers). Other examples
include remote terminals, kiosks, and video game machines (e.g.,
video game consoles, portable gaming devices, and the like). Other
examples include communication devices (including cellular phones,
such as smart phones), and media devices (including recorders,
editors, and players such as televisions, set-top boxes, music
players, digital photo frames, and digital cameras). Additionally,
the electronic device 100 could be a host or a slave to the sensor
102.
[0028] Sensor 102 can be implemented as a physical part of the
electronic device 100, or can be physically separate from the
electronic device 100. As appropriate, the sensor 102 may
communicate with parts of the electronic device 100 using any one
or more of the following: buses, networks, and other wired or
wireless interconnections. Examples include I.sup.2C, SPI, PS/2,
Universal Serial Bus (USB), Bluetooth, RF, and IRDA.
[0029] In some embodiments, sensor 102 will be utilized as a
fingerprint sensor utilizing one or more various electronic
fingerprint sensing methods, techniques and devices to capture a
fingerprint image of a user. In other embodiments, others type of
biometric sensors or input devices may be utilized instead of or in
addition to the fingerprint sensor to capture a biometric sample.
For instance, input devices that capture other biometric data such
as faces, vein patterns, voice patterns, hand writing, keystroke
patterns, heel prints, body shape, and/or eye patterns, such as
retina patterns, iris patterns, and eye vein patterns may be
utilized. For ease of description, biometric data discussed herein
will be in reference to fingerprint data. However, any other type
of biometric data could be utilized instead of or in addition to
the fingerprint data.
[0030] Generally, fingerprint sensor 102 may utilize any type of
technology to capture a user's fingerprint. For example, in certain
embodiments, the fingerprint sensor 102 may be an optical,
capacitive, thermal, pressure, radio frequency (RF) or ultrasonic
sensor. Optical sensors may utilize visible or invisible light to
capture a digital image. Some optical sensors may use a light
source to illuminate a user's finger while utilizing a detector
array, such as a charge-coupled device (CCD) or CMOS image sensor
array, to capture an image.
[0031] Regarding capacitive sensors, capacitive sensing
technologies include two types: passive and active. Both types of
capacitive technologies can utilize similar principles of
capacitance changes to generate fingerprint images. Passive
capacitive technology typically utilizes a linear one-dimensional
(1D) or a two-dimensional (2D) array of plates (i.e., electrodes or
traces) to apply an electrical signal, e.g., in the form of an
electrical field, such as a varying high speed (RF or the like)
signal transmitted to the finger of the user from a transmitter
trace and received at a receiver trace after passage through the
finger. A variation in the signal caused by the impedance of the
finger indicates, e.g., whether there is a fingerprint valley or
ridge between the transmitter trace and the receiver trace in the
vicinity of where the transmission and reception between the traces
occurs. Fingerprint ridges, as an example, can typically display
far less impedance (lower capacitance across the gap) than valleys,
which may exhibit relatively high impedance (higher capacitance
across the gap). The gaps can be between traces on the same plane,
horizontal, vertical or in different planes.
[0032] Active capacitive technology is similar to passive
technology, but may involve initial excitation of the epidermal
skin layer of the finger by applying a current or voltage directly
to the finger. Typically, thereafter, the actual change in
capacitance between the source of the voltage or current on an
excitation electrode (trace) and another receptor electrode (trace)
is measured to determine the presence of a valley or ridge
intermediate the source electrode and the another receptor
electrode.
[0033] In some embodiments of the capacitive sensor, the traces may
form a plurality of transmitter electrodes and a single receiver
electrode or a plurality of receiver electrodes and a single
transmitter electrode arranged in a linear one dimensional
capacitive gap array. In such embodiments, the capacitive gap may
be horizontal across the gap formed by the respective ends of the
plurality of traces and the single trace, whether transmitter or
receiver.
[0034] In some embodiments of the capacitive sensor, the traces may
form a 2D grid array, e.g., with rows of transmitter/receiver
traces on one substrate and columns of receiver/transmitter traces
on the same or a separate substrate, e.g., laminated together with
some form of dielectric between the traces to form a 2D sensor
element array. A 2D array may also be formed using a 2D matrix of
sensing electrodes. Such 2D arrays may form a 2D placement sensor
array (also sometimes known as an "area sensor" or "touch sensor")
or a 2D swipe sensor array (also sometimes known as a "slide
sensor"). A swipe sensor may also be formed from a one or more 1D
arrays or linear arrays.
[0035] Regarding thermal sensors, when a finger is presented to a
thermal sensor, the fingerprint ridges make contact with the sensor
surface and the contact temperature is measured. The ridges contact
the sensor and yield a temperature measurement, while the valleys
do not make contact and are not measured beyond some captured
ambient noise. A fingerprint image is created by the
skin-temperature of the ridges that contact the sensor and the
ambient temperature measure for valleys.
[0036] Regarding pressure sensors, there are two types of pressure
sensing detectors available, which include conductive film
detectors and micro electro-mechanical devices (MEMS). Conductive
film sensors use a double-layer electrode on flexible films.
Accordingly, a user who presses their finger to the flexible films
will leave an imprint that is utilized to capture an image of the
fingerprint. MEMS sensors use small silicon switches on a silicon
chip, such that when a fingerprint ridge touches a switch, it
closes and generates an electronic signal. The electronic signals
are detected and utilized to create an image of the fingerprint
pressed to the MEMS sensor.
[0037] Regarding RF sensors, a user's finger is pressed to the
sensor, which in turn applies an RF signal to the fingerprint
touched to the sensor. The fingerprint reflects a portion of the
applied RF signal which is in turn detected by a pixel array of the
sensor. The detected signal is utilized to create a fingerprint
image of the user's fingerprint.
[0038] Regarding ultrasonic sensors, these types of sensor utilized
very high frequency sound waves to penetrate an epidermal layer of
skin of a user's fingerprint pressed to the sensor. Typically, the
sound waves are generated using a piezoelectric transducer, which
also functions to receive the sound waves reflected from the user's
fingerprint. These reflected sound waves are detected and utilized
to create an image of the user's fingerprint.
[0039] Biometric image sensors, such as fingerprint sensors, such
as the sensor 102, which detect and measure features of the surface
of a finger using one or more of optical, capacitive, thermal,
pressure, RF and ultrasonic technologies, as discussed above,
sometimes fall into one of two categories: (1) placement sensors
and (2) typically smaller swipe sensors. Placement sensors have an
active sensing surface that is large enough to accommodate at least
a portion of the relevant part of the fingerprint of the finger
during a single scan or sensing action. Generally, the placement
sensors are rectangular in shape with a sensing surface area that
ranges from around 100 mm.times.100 mm down to 10 mm.times.10 mm,
or in some instances smaller than 10 mm.times.10 mm. Accordingly,
for small size placement sensors, only a portion of the fingerprint
will be captured either for immediate use in a recognition process
or as part of a fingerprint template for later use in the
recognition process. Additionally, in certain embodiments, the
placement sensor could have a non-rectangular shape and sensing
surface area. Typically, for placement sensors, the finger is held
stationary over the sensing surface during a measurement.
Generally, during a fingerprint enrollment process, multiple views
of the fingerprint image will be captured.
[0040] Generally, swipe sensors are smaller in size than placement
sensors and require the finger to be moved over the sensor during a
measurement. Typically, the finger movement will be either 1D in
that the finger moves in a single direction over the sensor
surface, or the finger movement can be 2D in that the finger can
move in more than one direction over the sensor surface during a
measurement. Generally, a fingerprint image captured during a
single frame will only be of a portion of a user's fingerprint, and
the sensor will capture a series of frames as the user swipes their
finger over the sensor so that a larger area of the fingerprint can
be captured in a single user input. The larger area may encompass a
full fingerprint, or it may still encompass only a partial
fingerprint, such as in a swipe sensor having a width less than the
full width of the finger.
[0041] Turning now to the processing system 104 from FIG. 1, basic
functional components of the electronic device 100 utilized during
capturing and storing a user fingerprint image are illustrated. The
processing system 104 includes a processor 106, a memory 108, a
template storage 110 and an operating system (OS) 112 hosting an
application suite 114 and a matcher 116. Each of the processor 106,
the memory 108, the template storage 110 and the operating system
112 are interconnected physically, communicatively, and/or
operatively for inter-component communications.
[0042] As illustrated, processor(s) 106 is configured to implement
functionality and/or process instructions for execution within
electronic device 100 and the processing system 104. For example,
processor 106 executes instructions stored in memory 108 or
instructions stored on template storage 110. Memory 108, which may
be a non-transitory, computer-readable storage medium, is
configured to store information within electronic device 100 during
operation. In some embodiments, memory 108 includes a temporary
memory, an area for information not to be maintained when the
electronic device 100 is turned off. Examples of such temporary
memory include volatile memories such as random access memories
(RAM), dynamic random access memories (DRAM), and static random
access memories (SRAM). Memory 108 also maintains program
instructions for execution by the processor 106.
[0043] Template storage 110 comprises one or more non-transitory
computer-readable storage media. The template storage 110 is
generally configured to store enrollment views for fingerprint
images for a user's fingerprint. The template storage 110 may
further be configured for long-term storage of information. In some
examples, the template storage 110 includes non-volatile storage
elements. Non-limiting examples of non-volatile storage elements
include magnetic hard discs, optical discs, floppy discs, flash
memories, or forms of electrically programmable memories (EPROM) or
electrically erasable and programmable (EEPROM) memories.
[0044] The processing system 104 also hosts an operating system
112. The operating system 112 controls operations of the components
of the processing system 104. For example, the operating system 112
facilitates the interaction of the processor(s) 106, memory 108 and
template storage 110. The operating system 112 further hosts the
application suite 114. The application suite 114 contains
applications utilizing data stored on the memory 108 or the
template storage 110 or data collected from interface devices such
as the sensor 102 to cause the processing system 104 to perform
certain functions. For instance, in certain embodiments, the
application suite 114 hosts an enroller application, which
functions to capture one or more views of the user's fingerprint.
The views or fingerprint images generally contain a partial or full
image of the user's fingerprint, and they may be raw images or
feature sets extracted from the raw images. The enrollment
application generally instructs the user to hold or swipe their
finger across the sensor 102 for capturing the image. After each
requested image is captured, the enrollment application typically
stores the captured image in the template storage 110. In certain
embodiments, the enrollment application will cause the data
representing the captured image to undergo further processing. For
instance, the further processing may be to compress the data
representing the captured image such that it does not take as much
memory within the template storage 110 to store the image.
[0045] In certain embodiments, the application suite 114 will also
contain applications for authenticating a user of the electronic
device 100. For example, these applications may be an OS logon
authentication application, a screen saver authentication
application, a folder/file lock authentication application, an
application lock and a password vault application. In each of these
applications, the individual application will cause the operating
system 112 to request the user's fingerprint for an authentication
process prior to undertaking a specific action, such as providing
access to the OS 112 during a logon process for the electronic
device 100. To perform this process, the above listed applications
will utilize the matcher 116 hosted by the operating system
112.
[0046] The matcher 116 of the operating system 112 functions to
compare the fingerprint image or images stored in the template
storage 110 with a newly acquired fingerprint image or images from
a user attempting to access the electronic device 100. In certain
embodiments, the matcher 116, or other process, will further
perform image enhancement functions for enhancing a fingerprint
image. An example of the image enhancement function is illustrated
in FIGS. 2a and 2b. FIG. 2a illustrates an unenhanced fingerprint
image that shows various ridges and minutiae of a fingerprint. As
can be seen in FIG. 2a, the image is noisy such that portions of
the image are cloudy and the ridges or contours are broken. FIG. 2b
illustrates the same fingerprint after the matcher 116 has
performed the image enhancement function. As can be seen, the image
enhancement function removes much of the noise such that the image
is no longer cloudy and the ridges are no longer broken.
[0047] In certain embodiments, the matcher 116, or other process,
is also configured to perform feature extraction from the
fingerprint image or images of the user. During feature extraction,
the matcher 116 will extract unique features of the user's
fingerprint to derive a verification template used during matching.
Various discriminative features may be used for matching,
including: minutia matching, ridge matching, ridge flow matching,
or some combination thereof. If authentication is performed using
minutia features, the matcher 116 will scan the captured view of
the user's fingerprint for minutia. FIG. 3 illustrates various
types of fingerprint minutia, including, from left to right, a
bridge point between two or more ridges, a dot, an isolated ridge,
an ending ridge, a bifurcation point and an enclosure. During
extraction, the matcher 116 acquires a location and orientation of
the minutia from the fingerprint and compares it to previously
captured location and orientation information of minutia from the
fingerprint image or images in the template storage 110.
[0048] The matcher may compare the verification template to the
enrollment template to compute a composite match score between the
templates. If the composite match score satisfies a threshold, the
match 116 indicates a match. Otherwise, a non-match may be
indicated.
[0049] In embodiments of the invention, an enrollment template may
contain multiple enrolled views of the user's fingerprint. Each of
the enrolled views may be stored separately, along with the
geometric relationships between the views. Alternatively, instead
of pre-computing the geometric relationships between the enrollment
views and storing them in the enrollment template, the geometric
relationships can be computed at match time prior to each match
attempt. The matcher 116 may calculate a composite match score
between the verification template and the enrollment template based
on individual match statistics derived from comparing a
verification view of the verification template with each individual
enrollment view in the enrollment template, or some subset of the
individual enrollment views in the enrollment template. The
individual match statistics may be collectively analyzed, along
with inter-view match statistics between enrollment views within
the enrollment template to calculate a composite match score. The
inter-view match statistics between the enrollment views may be
pre-computed and stored in the enrollment template, or may be
computed at match time.
[0050] The composite match score may be calculated as a function of
both the individual match statistics and the inter-view match
statistics. For example, individual match scores may be computed
based on comparing the verification view with a plurality of
individual enrollment views in the enrollment template. Instead of
simply combining these individual scores, the matcher may combine
these scores along with some adjustment based on the inter-view
match statistics between individual enrollment views. As another
example, feature vectors from the inter-view match statistics and
individual match statistics, such as transformation errors, number
of predicted matches, number of non-predicted matches, etc. can be
derived from the relationships between enrollment views and fed
into a neural network or other machine learning classifier to
calculate an overall composite match score.
[0051] FIG. 4 illustrates an embodiment of the matcher 116 of FIG.
1, which uses a machine learning classifier to calculate a
composite match score. In the embodiment illustrated in FIG. 4, the
matcher 116 includes a matching module 402 that utilizes a
fingerprint enrollment template 404 for a recognition process. As
used herein, a recognition process includes authentication or
verification processes, which are utilized to verify a user by
determining whether a match is found with confidence between a
biometric sample captured by an input device and an enrollment
template previously captured from the user. Additionally, the
recognition process could also pertain to an identification
process, which functions to identify a user based on the user's
verification template being compared to a database containing a
plurality of enrollment templates.
[0052] The fingerprint enrollment template 404 includes a plurality
of fingerprint views of a user's fingerprint. In the illustrated
embodiment, the fingerprint enrollment template 404 includes four
enrollment views, E.sub.1, E.sub.2, E.sub.3 and E.sub.4. In the
illustrated embodiment, the recognition process is being used for a
user fingerprint view 406. The matching module 402 determines match
statistics between the fingerprint enrollment template 404 and the
fingerprint verification view 406. The match statistics may include
individual match statistics computed by comparing the verification
view 406 to each of the individual enrollment views 404. The match
statistics are provided to a machine learning module 408, which in
turn calculates a composite multi-view match score 410. The
composite multi-view match score 410 provides an indication of
confidence of a match between the verification fingerprint view 406
and the fingerprint enrollment template 404.
[0053] In one embodiment, the machine learning module uses a neural
network to calculate the composite score. However, the machine
learning module may use other scoring methods, such as support
vector machines, random forests, regression models, and the
like.
[0054] The machine learning module 408, in addition to the
individual match statistics determined by the matching module 402,
in certain embodiments, utilizes inter-view match statistics to
calculate the composite multi-view match score 410. The inter-view
match statistics represent match statistics between enrollment
views of the fingerprint enrollment template 404.
[0055] FIG. 5 illustrates a geographical layout of the enrollment
views E.sub.1, E.sub.2, E.sub.3 and E.sub.4 and the verification
fingerprint view 406 within a fingerprint image boundary 500,
according to an example embodiment. The fingerprint image boundary
represents a total surface area a user's fingerprint image, when
captured, may encompass. In this example, each enrollment view and
the verification fingerprint view 406 represent only partial images
of the overall fingerprint 500. The portion of the fingerprint
captured in the enrollment views and the verification fingerprint
view 406 will be determined by how a user interacts with the sensor
102 (see FIG. 1). In this regard, the enrollment views and the
verification fingerprint view 406 may be geographically dispersed
over the potential surface area of the fingerprint represented by
the fingerprint image boundary 500. As illustrated, the enrollment
views and the verification fingerprint view 406 represent portions
of the fingerprint and therefore will contain discriminative
features such as ridges and valleys, ridge flows, and/or
fingerprint minutia, etc., as illustrated in FIG. 3.
[0056] The individual match statistics and/or inter-view match
statistics that are collected can vary in different
implementations.
[0057] In certain embodiments, the match statistics, including the
individual match statistics between the verification view and the
individual enrollment views, the inter-view match statistics
between the enrollment views, or both, takes into account any
geometric relationship between the boundaries of the enrollment
views and the verification fingerprint view 406. For instance, the
geometric relationship may include a geometric transformation
between the enrollment views and the verification fingerprint view
406. The transformation may be performed in any one or more
coordinate systems, including, but not limited to a Cartesian
coordinate system, a polar coordinate system, a spherical
coordinate system or a cylindrical coordinate system. For example,
in an embodiment relying on the Cartesian coordinate system, the
geometric transformation would include a translation in the x
direction "Dx" and/or a translation in the y direction "Dy."
[0058] Additionally, the enrollment views and the verification
fingerprint view 406 are illustrated as just being boxes; however,
each box will contain a portion of a fingerprint that depending on
how the user interacts with the sensor 102 (see FIG. 1) may not
only be of a different portion of the fingerprint but also may have
a different rotation from the other boxes. Therefore, an additional
point for collection as a match statistic may be a rotation Dtheta
between each of the enrollment views and the verification
fingerprint view 406.
[0059] Furthermore, as mentioned, each of the enrollment views and
the verification fingerprint view 406 may contain certain minutia
points. In embodiments relying on minutia features, a number of
matched minutia points "k" may be utilized as part of the match
statistics.
[0060] In certain embodiments, the individual match statistics
and/or inter-view match statistics include a match score. The
individual or inter-view match score may be determined by comparing
the views according to some matching metric between features in the
views. If minutia-based matching is used, for example, the match
score may be correlated to a number of matching minutia between the
views, which in turn may be correlated to an amount of overlap
between compared views. Other scoring methods may exhibit similar
correlation between amount of overlap and match score. Thus, in
some implementations, a greater amount of overlap resulting from
the geometric transformation between views will correspond to a
stronger match score, and similarly a lower amount of overlap
between views will correspond to a weaker match score.
[0061] In some embodiments, predicted weak matches, strong matches,
or non-matches can be derived from an individual geometric
transformation of the verification view and inter-view geometric
transformations between the enrollment views. Accordingly, in the
illustrated embodiment, the verification fingerprint view V
overlaps strongly with both E.sub.2 and E.sub.3 overlaps minimally
with both E.sub.1 or E.sub.4. Therefore, a strong match score
should exist for the match statistics between the verification
fingerprint view V and E.sub.2 and E.sub.3, while the match score
between the verification fingerprint view V and E.sub.1 and E.sub.4
should be low. In this example, these match statistics can be
predicted and the composite match score can be adjusted based on
consistency between these predicted match scores and the actually
computed individual match scores. For example, the inter-view
geometric transformations among E.sub.1, E.sub.2, E.sub.3, and
E.sub.4, and the individual geometric transformation of V to
E.sub.2, indicates that one strong match and two weak matches
should be expected when computing the individual match scores for V
against E.sub.1, E.sub.3, and E.sub.4. If the actual individual
match scores are consistent with this prediction, e.g., one strong
match and two weak matches are determined, the composite score may
include an increase due to this high consistency. Alternatively, if
the actual individual match scores deviate significantly from this
prediction, the composite match score may decrease.
[0062] As mentioned previously, all of the above mentioned match
statistics are not only determined for the enrollment views
compared against the verification fingerprint view 406 but also
between each of the enrollment views E.sub.1, E.sub.2, E.sub.3 and
E.sub.4. This collection 600 of match statistics is geometrically
illustrated in FIG. 6. In FIG. 6, the solid lines represent
inter-view match statistics between individual enrollment views in
the fingerprint enrollment template 404 (see FIG. 4), and the
dashed lines represent match statistics between the verification
fingerprint view 406 and the individual enrollment views in the
fingerprint enrollment template 404. In some embodiments, the match
statistics collected include a match score, a geometric translation
in the x-direction Dx, a geometric translation in the y-direction
Dy, a rotation Dtheta, the number of matched minutia k, and/or a
number of non-matched minutia. Other match statistics can be
collected, and the specific set of match statistics collected and
used to calculate the composite match score can vary from
embodiment to embodiment.
[0063] Based on a portion or all of the acquired match statistics
illustrated in FIG. 6, the composite multi-view match score 410
(see FIG. 4) can be determined. By utilizing the match statistics
represented by the dashed lines in FIG. 6, the matcher 116 (see
FIG. 1) will be able to improve the reliability of the calculation
of the composite multi-view match score 410.
[0064] FIGS. 7-8 depict examples of composite match score
calculations using consistency checks between the individual match
statistics and the inter-view match statistics.
[0065] FIGS. 7(a)-7(c) depicts an example in which matched minutia
in the individual match statistics and inter-view match statistics
are compared, and the composite match score is adjusted based on
correspondence therebetween. As shown in FIGS. 7(a)-7(c),
verification view V1, and individual enrollment views E1 and E2 all
overlap and contain common minutia. FIG. 7(c) shows the match
statistics for these minutia (a.sub.i, b.sub.i, c.sub.i) each of
the pairings of views, including both match minutia (e.g., a.sub.2
and c.sub.3 match in the pairing of E1 to V1), and non-matching
minutia (e.g., neither a.sub.1 or c.sub.3 match any minutia in the
pairing of E1 to V1). As a simplified example, since a.sub.4
matches c.sub.6 in the individual pairing of E1 to V1 and a.sub.4
also matches b.sub.1 in the inter-view pairing of E1 to E2, c.sub.6
should match b.sub.1 in the individual pairing of V1 to E2. The
individual match statistics can be compared to the inter-view match
statistics regarding match minutia, and the composite match score
can be adjusted accordingly. In this example, the composite score
can be increased when the minutia in V1 matches the results from
the other statistics.
[0066] FIG. 8 depicts an example in which a transformation error is
computed by deriving a geometric transformation from the inter-view
match statistics, and comparing this to an individual geometric
transformation determined by comparing the verification view and an
individual enrollment view (e.g., using a standard image alignment
technique). In this example, V1 is associated with E1 by
transformation T(V1:E1). E1 is associated with E2 by geometric
transformation T(E1:E2). From this information, a transformation
between V1 and E2 can be derived from these relationships. To
compute a transformation error, an individual geometric
transformation T'(V1:E2) can be separately computed by direct
comparison of V1 to E2 using a standard image alignment technique.
The computed transformation and derived transformation can be
compared to generate a transformation error. The composite match
score can then further be a function of this comparison. For
example, the composite match score can be increased or decreased in
relation to this transformation error (where a high error or large
difference between the transformations indicated inconsistencies
that may result in a decrease in the composite match score and vice
versa). Alternative, the derived transformations or the
transformation error may be used as a feature vector input into a
machine learning classifier used to compute a composite match
score.
[0067] FIG. 9 illustrates a flow chart 700 for a recognition
process performed by the matcher 116 of FIGS. 1 and 4, in
accordance with an embodiment of the invention. During an
enrollment process (not pictured), the matcher 116 acquires the
fingerprint enrollment template 404. The fingerprint enrollment
template 404 is captured, typically at some point in time prior to
performing the recognition process, and stored either locally at
the device 100 (see FIG. 1) or externally and uploaded to the
device at a later point in time. The enrollment template can be
derived from multiple biometric samples captured by a fingerprint
sensor or other input device, and may include multiple
corresponding enrollment views.
[0068] At step 704, the matcher 116 (see FIG. 1) acquires a
verification template 406 (see FIG. 4) for use in the recognition
process. The verification template includes a verification view of
a biometric sample (e.g., a sensed fingerprint), and includes a raw
image or a feature set extracted from a raw image. The individual
fingerprint view 406 will typically be collected at the time of
performing the recognition process to be compared against the
previously collected fingerprint enrollment template 404. However,
in certain embodiments, the individual fingerprint view 406 may
have been previously captured and uploaded to the device for use in
the recognition process.
[0069] At step 706, the matcher 116 (see FIG. 1) compares the
verification view to a plurality of individual enrollment views in
the enrollment template to determine individual match statistics
between the verification view and each of the individual enrollment
view 406 (see FIG. 4) and each view of the fingerprint enrollment
template 404. As mentioned previously, in a certain embodiment,
these individual match statistics can include a geometric
translation, a rotation, and an individual match score. The
individual match statistics may be determined against all of the
enrollment views in the enrollment template, or only some subset of
them.
[0070] At step 708, the matcher 116 (see FIG. 1) utilizes the
individual match statistics to calculate a composite multi-view
match score 410 (see FIG. 4). The composite match score is computed
as a function of the individual match statistics and the inter-view
match statistics as described above. For example, the composite
match score may be a combination of the individual match scores
from the previous step, with adjustments based on the inter-view
match statistics. Alternatively, it may be calculated using a
machine learning scoring module that received the individual match
statistics and inter-view match statistics as inputs.
[0071] At step 710, the matcher 116 (see FIG. 1) indicates whether
a match has been found based on the composite multi-view match
score 410. In certain embodiments, this indication is based on a
comparison between the composite multi-view match score 410 and a
threshold value. If the composite multi-view match score 410 is
above the threshold, then a match is indicated with a high level of
confidence; however, if the composite multi-view match score 410 is
below the threshold, then a non match is indicated.
[0072] All references, including publications, patent applications,
and patents, cited herein are hereby incorporated by reference to
the same extent as if each reference were individually and
specifically indicated to be incorporated by reference and were set
forth in its entirety herein.
[0073] The use of the terms "a" and "an" and "the" and "at least
one" and similar referents in the context of describing the
invention (especially in the context of the following claims) are
to be construed to cover both the singular and the plural, unless
otherwise indicated herein or clearly contradicted by context. The
use of the term "at least one" followed by a list of one or more
items (for example, "at least one of A and B") is to be construed
to mean one item selected from the listed items (A or B) or any
combination of two or more of the listed items (A and B), unless
otherwise indicated herein or clearly contradicted by context. The
terms "comprising," "having," "including," and "containing" are to
be construed as open-ended terms (i.e., meaning "including, but not
limited to,") unless otherwise noted. Recitation of ranges of
values herein are merely intended to serve as a shorthand method of
referring individually to each separate value falling within the
range, unless otherwise indicated herein, and each separate value
is incorporated into the specification as if it were individually
recited herein. All methods described herein can be performed in
any suitable order unless otherwise indicated herein or otherwise
clearly contradicted by context. The use of any and all examples,
or exemplary language (e.g., "such as") provided herein, is
intended merely to better illuminate the invention and does not
pose a limitation on the scope of the invention unless otherwise
claimed. No language in the specification should be construed as
indicating any non-claimed element as essential to the practice of
the invention.
[0074] Preferred embodiments of this invention are described
herein, including the best mode known to the inventors for carrying
out the invention. Variations of those preferred embodiments may
become apparent to those of ordinary skill in the art upon reading
the foregoing description. The inventors expect skilled artisans to
employ such variations as appropriate, and the inventors intend for
the invention to be practiced otherwise than as specifically
described herein. Accordingly, this invention includes all
modifications and equivalents of the subject matter recited in the
claims appended hereto as permitted by applicable law. Moreover,
any combination of the above-described elements in all possible
variations thereof is encompassed by the invention unless otherwise
indicated herein or otherwise clearly contradicted by context.
* * * * *